Skip to content

Base Contrast Axis

The Base axis represents the fundamental tissue contrast weighting of an MRI series. It answers the question: "What physical contrast mechanism is this image based on?"


Overview

Base contrast is determined by the MRI physics parameters (TR, TE, TI) and acquisition type. Each base contrast emphasizes different tissue properties:

Base Full Name Primary Signal Clinical Use
T1w T1-weighted Short TR/TE, WM bright Anatomy, Gd enhancement
T2w T2-weighted Long TR/TE, fluid bright Pathology detection
PDw Proton density Long TR, short TE Tissue composition
T2*w T2-star weighted GRE, susceptibility Iron, blood products
DWI Diffusion-weighted Water motion Stroke, tumors
PWI Perfusion-weighted Blood flow Hemodynamics
SWI Susceptibility-weighted Phase + magnitude Veins, microbleeds
MTw Magnetization transfer Bound protons Myelin integrity
T1rho T1-rho weighted Spin-lock Cartilage, research

Detection Strategy

NILS uses a four-tier detection approach (in priority order):

Tier 1: Technique Inference

Some pulse sequences physics-lock to a specific base contrast:

Technique Implied Base Confidence
MPRAGE, MEMPRAGE, MP2RAGE T1w 95%
TOF-MRA T1w 90%
ME-GRE, comb-ME-GRE T2*w 85-90%
SWI SWI 95%
DWI-EPI DWI 95%
ASL-EPI, DSC-EPI PWI 90-95%

Why? These techniques have physics constraints that determine contrast. MPRAGE always has T1-weighting due to its IR-prepared 3D GRE design.

Tier 2: Exclusive Flags

Definitive signals from unified_flags that immediately identify base:

DWI indicators: - is_dwi - Diffusion sequence flag - has_adc - ADC map present - has_fa - FA map present - has_trace - Trace diffusion

PWI indicators: - is_perfusion - Perfusion flag - is_asl - Arterial spin labeling - has_cbf, has_cbv, has_mtt, has_tmax, has_ttp - Perfusion maps

SWI indicators: - is_swi - SWI sequence - has_swi - SWI processing

MTw indicator: - has_mtc - Magnetization transfer contrast

Synthetic MRI outputs: - has_t1_synthetic → T1w - has_t2_synthetic → T2w - has_flair_synthetic → T2w (FLAIR base) - has_pd_synthetic → PDw

Tier 3: Keyword Matching

Text search in series description for explicit contrast labels:

Keywords Base
t1w, t1 weighted, t1-w T1w
t2w, t2 weighted, t2-w T2w
pdw, proton density PDw
t2*, t2star T2*w
dwi, diffusion, dti DWI
pwi, perfusion, dsc, dce PWI
swi, swan, venobold SWI
mtw, magnetization transfer MTw

Tier 4: Physics-Based Inference

When other methods fail, use TR/TE/TI thresholds (validated against 400K+ fingerprints):

Spin Echo (SE) Family

Contrast TR TE Notes
T1w < 1000ms < 30ms Short TR/TE
T2w > 2000ms > 50ms Long TR/TE
PDw > 2000ms < 30ms Long TR, short TE

Gradient Echo (GRE) Family

Contrast TE Notes
T1w < 10ms Short TE
T2*w > 15ms Long TE for susceptibility

Inversion Recovery (IR) Family

Contrast TI Notes
T1w (standard IR) 300-1500ms Medium TI
T2w (STIR-like) < 300ms Short TI nulls fat

Special Cases

FLAIR Differentiation

FLAIR can be T1-FLAIR or T2-FLAIR, differentiated by TE:

Type TE Threshold Base Typical Use
T1-FLAIR < 40ms T1w Research
T2-FLAIR ≥ 40ms T2w Clinical (most common)

Database validation shows: - 97% of series with "t1" keyword + FLAIR have TE < 40ms - 99.9% of series with "t2" keyword + FLAIR have TE ≥ 80ms

FLAIR Modifier

The FLAIR modifier is detected on the Modifier axis separately. Here we only determine the underlying base contrast (T1w vs T2w).

Dual-Echo PD+T2

Dual-echo sequences acquire both PD and T2 in a single acquisition. NILS splits these into separate stacks and uses TE to determine which echo:

Echo Type TE Base
PD echo < 40ms PDw
T2 echo ≥ 40ms T2w

Detection triggers when BOTH "pd" and "t2" keywords appear in the series description.

MP2RAGE Outputs

MP2RAGE is fundamentally a T1-weighted technique. All outputs receive T1w base:

Output Flags Base
INV1 is_mp2rage_inv1 T1w
INV2 is_mp2rage_inv2 T1w
UNI has_uniform T1w
UNI-DEN is_uniform_denoised T1w
T1 Map has_t1_map T1w

Quantitative Maps

Relaxometry maps have an implied base from their measurement:

Map Base Rationale
T1 Map, R1 T1w Measures T1 relaxation
T2 Map, R2 T2w Measures T2 relaxation

Primary Tissue Contrasts

T1w (T1-Weighted)

Physics: Short TR (~500-800ms) allows incomplete T1 recovery, emphasizing T1 differences. Short TE (~10-20ms) minimizes T2 effects.

Signal characteristics: - White matter: bright (short T1) - Gray matter: intermediate - CSF: dark (long T1) - Fat: bright

Clinical applications: - Anatomy reference - Gadolinium enhancement - Post-contrast lesion detection

T2w (T2-Weighted)

Physics: Long TR (>2000ms) allows full T1 recovery, eliminating T1 contrast. Long TE (>50ms) emphasizes T2 differences.

Signal characteristics: - White matter: dark - Gray matter: intermediate - CSF: very bright (long T2) - Edema: bright

Clinical applications: - Pathology detection - MS lesions - Tumor visualization - Infection/inflammation

PDw (Proton Density)

Physics: Long TR (>2000ms) removes T1 contrast. Short TE (<30ms) minimizes T2 contrast. Signal primarily reflects hydrogen density.

Signal characteristics: - Intermediate contrast - CSF: bright - Tissues differentiated by proton content

Clinical applications: - Often paired with T2w (dual-echo) - Cartilage imaging - Research applications

T2*w (T2-Star Weighted)

Physics: GRE sequence without 180° refocusing pulse. Sensitive to microscopic field inhomogeneities (susceptibility effects).

Signal characteristics: - Blood products: dark (paramagnetic) - Iron deposits: dark - Calcium: dark - Air-tissue interfaces: signal dropout

Clinical applications: - Hemorrhage detection - Microbleed identification - Iron quantification - Basis for SWI and BOLD


Specialized Contrasts

DWI (Diffusion-Weighted)

Physics: Measures random motion of water molecules using diffusion gradients. Contrast depends on b-value.

Signal characteristics: - Restricted diffusion: bright (acute stroke) - Free water: dark - High b-value → more diffusion weighting

Derived maps: - ADC (Apparent Diffusion Coefficient) - FA (Fractional Anisotropy) - MD (Mean Diffusivity) - Trace

PWI (Perfusion-Weighted)

Physics: Measures blood flow dynamics. Multiple methods: - DSC: Dynamic susceptibility contrast (T2-based) - DCE: Dynamic contrast-enhanced (T1-based) - ASL*: Arterial spin labeling (non-contrast)

Outputs: - CBF (Cerebral Blood Flow) - CBV (Cerebral Blood Volume) - MTT (Mean Transit Time) - Tmax, TTP (time parameters)

SWI (Susceptibility-Weighted)

Physics: Combines GRE magnitude and filtered phase images to enhance susceptibility contrast.

Signal characteristics: - Paramagnetic substances: very dark - Veins: dark (deoxyhemoglobin) - Superior to T2* for small structures

Applications: - Microbleed detection - Venous imaging - Iron quantification - QSM derivation

MTw (Magnetization Transfer)

Physics: Off-resonance RF pulse saturates bound (macromolecular) protons. Saturation transfers to free water pool, reducing signal in tissues with bound protons.

Signal characteristics: - Myelin-rich tissue: signal loss - Free water: minimal effect

Applications: - MS lesion characterization - Myelin integrity assessment


Confidence Levels

Detection confidence varies by method:

Method Confidence Rationale
Technique inference 95% Physics-locked
Exclusive flags 90% Definitive markers
Keywords 85% Text-based
FLAIR/Dual-echo TE 85% Physics threshold
Physics ranges 70% Edge case fallback
Unknown fallback 50% Could not determine

Conflict Detection

NILS checks for conflicts between detection method and text evidence:

  • If physics predicts T1w but text contains "t2w" → conflict flagged
  • Conflicts trigger manual review
  • Authoritative methods (technique inference, exclusive flags) skip conflict check

YAML Configuration

Base detection is configured in backend/src/classification/detection_yaml/base-detection.yaml:

bases:
  T1w:
    name: "T1w"
    description: "T1-weighted; short TR/TE, white matter bright"
    keywords:
      - "t1w"
      - "t1 weighted"
    physics:
      se:
        tr_max: 1000
        te_max: 30
      gre:
        te_max: 10

technique_inference:
  MPRAGE: ["T1w", 0.95]
  MP2RAGE: ["T1w", 0.95]
  # ...

rules:
  allow_multiple: false  # Only one base per series
  priority_order:
    - "DWI"
    - "PWI"
    - "SWI"
    # ... (specialized first, then primary)